Genomic selection for feed efficiency and reduced environmental impact in dairy cattle

A computer simulation software that could simulate structured dairy cattle breeding schemes, with multiple genetic levels such as elite-breeding level, production level, and more countries, i.e. across country selection and use of data. The program can handle several traits such as milk production and feed intake. The amount of DNA data could be varied from low to high density genome-wide SNP (single nucleotide polymorphisms) up to whole genome sequence. With this software, alternative genomic selection and traditional breeding schemes were compared for the accuracy of selection and the rate of genetic improvement of milk production and feed efficiency in simulated breeding schemes that mimicked the Norwegian Red Cattle (NRC) population. When including feed efficiency in genomic selection schemes, high selection accuracies and more genetic progress was attained than when traditional selection was applied. However, the extra genetic gain for feed efficiency came at the cost of a reduction in genetic progress for milk production. Nonetheless, total genetic progress was higher when feed efficiency was included in the selection scheme. The required number of cows with records for feed intake was large and estimated at 4000 new cows with feed intake records annually. Mid-infrared (MIR) spectroscopy of milk was used to predict dry matter intake (DMI) and net energy intake (NEI). MIR milk-spectra data are widely available on Norwegian cows because they are used to predict protein and fat content in milk in Kukontrollen. Thus, successful prediction of DMI and NEI using MIR implies that we can predict the feed efficiency of all Norwegian cows that are in kukontrollen. Predictions were performed using either Partial Least Squares (PLS) or Best Linear Unbiased Prediction (BLUP). The greatest accuracy of prediction of DMI (54%) and NEI (65%) was achieved when including the MIR spectra and body-weight and milk yield in the prediction equation. Thus, MIR spectra data could be used to predict feed efficiency in Norwegian Red Cattle on a national scale and that the prediction accuracy is increased when also milk yield and body weight are included in the prediction equation. We subsequently used MIR on 134,880 milk samples on 35,600 cows to develop predictions of breeding values for net energy intake. First, the MIR data were used to predict daily net energy intake records. Next, breeding value evaluation methodology was developed based on the ?test day model?, which is a longitudinal data analysis that models the NEI records of the cows along the lactation. Here, only records from the first lactation were used, but extension to more lactations is straightforward. Also extension towards the inclusion of accurate actual feed intake records is straightforward using multitrait breeding value prediction theory. Currently, however, accurate recordings of feed intake are not available in sufficient quantities to be of use in breeding value estimation. However, the Geno Project ?Hightech Fjøs? (www.geno.no) intents to provide such data. In an international collaboration, Irish (Moorepark, Cork, Ireland) and Norwegian spectroscopy data of milk were combined to predict net energy intake (NEI) and the Irish effective energy intake (EEI). Partial Least Squares (PLS) and Best Linear Unbiased Predictions (BLUP) were implemented either as a single trait (sBLUP) or a multi-trait (mBLUP) method. Combining the spectra from two countries, increased the accuracy of prediction of EEI by 0.02 units in both the cross-validation and the external validation compared to the model with spectral information from one country only. Whereas for NEI, combining the spectra from two countries hardly affected the external validation accuracy. Thus, MIR spectral data from two countries can be combined and used to predict energy intake as a measure of feed intake in dairy cattle. We also developed methodology to make records across experimental farms comparable, which is needed to combine feed intake recordings from different experimental across and within countries.

Norwegian food production has to increase by about 1% per year until the year 2030 (Landbruksmeldinga, 2012). At the same time GHG emissions need to reduce and use of Norwegian food resources needs to increase. In dairy production, all these goals are add ressed by improving the feed efficiency of cows. In addition, the feeding costs comprise 70% of all variable costs, and improving feed efficiency can easily save 140 mNOK per year. Genomic selection is a novel selection method based on DNA information, wh ich is recently implemented for the selection of NRF cattle. This selection method is especially useful for traits which are impossible or too expensive to record on a large scale in practice. Feed efficiency and GHG emissions of lactating cows are very e xpensive to record on cows, and impossible to record on elite bulls, which are the most important breeding animals in dairy cattle. The project will thus develop genomic selection strategies for the genetic improvement of feed efficiency and GHG emissions . In order to maximise the accuracy of the genomic selection, we will combine data from experimental herds, from milking robots and automatic feeding stations, from approximate recordings of feed efficiency, which can be obtained from Kukontrollen, and da ta from other Nordic countries. The project is embedded in a larger Nordic project entitled 'Towards more sustainable dairy production in the Nordic countries through improved Feed Efficiency and reduced environmental impact', which was applied for at the 'Nordic Dairy Cattle R&D (NDC)' and was given 'highest priority' by NDC. Thus, the project exploits research synergies across the Nordic countries, involving Danish, Swedish and Finnish researchers, and across the disciplines animal genetics and animal n utrition in order to improve feed efficiency reduce environmental impacts.